Vibe Coding Purity Test
Check every item you've done.
- Posted a hot take thread about "SWE is dead" and meant it.
- Argued that "agents are just scripts" in replies.
- Said "prompts are the new programming language" unironically.
- Predicted the next model release month in public.
- Tweeted "evals are fake, vibes are real".
- Posted "context window is all you need".
- Dunked on a benchmark screenshot for being cherry-picked.
- Picked a side in open-source vs closed-model wars.
- Quoted-tweeted an AI doom thread with "skill issue".
- Posted a graph about exponentials to win an argument.
- Called a model launch a "nothing-burger" within an hour.
- Said "alignment is just product safety" to stir discourse.
- Started a "agents will replace PMs" debate.
- Tweeted "MCP is the new plugin ecosystem".
- Wrote "hot take: fine-tuning is dead".
- Complained about "AI slop" flooding the timeline.
- Ran a "doomer vs accelerationist" poll.
- Shared a "we're so back / it's so over" meme for a release.
- Claimed "RLHF ruined creativity" in a thread.
- Posted a prediction-market screenshot about AGI timelines.
- Made a "prompt injection will kill us all" take.
- Replied "benchmarks don't matter, users do".
- Declared a model "AGI-ish" and got ratioed.
- Live-tweeted a keynote to score points.
- Announced you're leaving AI Twitter, then returned next week.
- Tried ChatGPT for the first time
- Asked an LLM to explain a bug in your code
- Used AI to rewrite an email or message
- Copied a prompt from someone else and ran it
- Turned on AI autocomplete in your editor
- Installed Cursor just to test AI edits
- Used GitHub Copilot to finish a small script
- Asked Claude to summarize a long article or PDF
- Generated a README with AI
- Used AI to brainstorm product or feature names
- Used AI to turn notes into a checklist
- Let an AI refactor a function for readability
- Ran a "fix lint errors" prompt
- Generated unit tests with AI
- Used AI to draft a cold outreach message
- Tried v0 or similar to mock a simple UI
- Tried Replit Agent on a toy project
- Asked AI to compare two libraries
- Used AI to debug a Dockerfile
- Let AI write a SQL query for you
- Asked AI to explain MCPs in plain English
- Followed an AI demo video step-by-step
- Used AI to translate text into another language
- Asked AI for a study plan to learn a tech stack
- Generated a logo or image with AI
- Paid for Cursor Pro to lift composer limits
- Upgraded to Claude Pro the day you hit the free cap
- Bought ChatGPT Plus just for GPT-4 access
- Subscribed to Copilot just for inline completions
- Used Cursor Composer to rewrite an entire feature
- Paid for v0 to ship a landing page overnight
- Used Bolt to scaffold a full-stack app in an hour
- Let Replit Agent set up a database and deploy it
- Tried Windsurf because someone said it beats Cursor
- Applied to the Devin waitlist and checked email daily
- Ran Clawdbot or Moltbot on a repo-wide refactor
- Bought a second AI tool subscription in the same week
- Cancelled a model subscription after a price hike
- Blew through $100 in API credits in a weekend
- Set an API budget and still blew past it
- Had an AI API bill bigger than your phone bill
- Paid for a vector database plan for embeddings
- Rented a GPU on RunPod or Vast for local models
- Bought new hardware just to run a local model
- Paid for Vercel Pro to host an AI demo
- Bought a domain just for an AI side project
- Paid for a prompt marketplace template pack
- Spent $500 plus in a month on AI tooling
- Stacked three AI subscriptions at once
- Paid for an agent framework course to speed up work
- Explained to a partner what "agentic" means, twice.
- Made a shared prompt doc for your team and gave it a name.
- Turned a Jira ticket into a prompt and called it a spec.
- Asked an LLM to write your standup update.
- Started your day by asking a model for a plan.
- Kept a context.md file just for the model.
- Gave a model a "project tour" before asking for changes.
- Used a prompt to draft meeting notes from a call.
- Sent a friend a prompt template instead of explaining the idea.
- Explained "context window" at a family dinner.
- Told your team you were "pairing with Claude" today.
- Created a "prompt cookbook" and actually used it.
- Let an agent open a PR and wrote the description for you.
- Ran a "plan -> execute -> reflect" loop more than once in a day.
- Set up a Slack channel just for prompts.
- Shared a screen to show someone a Cursor flow.
- Asked a coworker to review a prompt like code.
- Apologized for a late reply because your agent was running.
- Added a "prompting guide" to your team wiki.
- Built a tiny script to feed repo context into a model.
- Wrote a "prompt contract" for how your team asks for changes.
- Explained why your "pair" does not take coffee breaks.
- Saved a model's system prompt as a permanent file.
- Used an LLM to triage bugs before the team did.
- Changed how you write commit messages because a model does the first pass.